 |
Description  |
|
|
TECHNICAL FIELD
This invention relates to robots and robotics used in the inspection of
parts.
BACKGROUND OF THE INVENTION
The automated factory of the future, in order to provide flexible
manufacturing of various types of parts, will require versatile robotic
type machines for inspecting the parts and for producing feedback
correction signals to the machines as well for insuring that defective
products are not released. In order to provide such inspection, various
programmable coordinate measuring machines have been proposed using
contact feelers and the like.
The weak links of virtually all of the approaches heretofore are generally
two. First, these approaches are only capable of providing dimensional
measurements and cannot inspect for the other aspects of the part, such as
defective surface finish. Second, these approaches are also slow because
of the time required for the sensor probe to make contact. It will be
appreciated that there are definite limits on the maximum speed at which a
measurement can be made when it is necessary to stop and touch a part in
order to make the measurement. This is particularly true when accuracies
that approach values as small as 2 or 3 microns are desired as in most
inspections of a finished machined part.
A general solution to the problems discussed in the previous paragraph
involves the use of an inspection robot based on optical sensing
principles. One example of a robot control system of this type is that
disclosed in my copending application Ser. No. 262,492, filed on May 11,
1981 and entitled "Electro-Optical Systems for Control of Robot
Manipulator Arms and Co-ordinate-Measuring Machines" now U.S. Pat. No.
4,453,085.
SUMMARY OF THE INVENTION
In accordance with the present invention, methods and apparatus are
provided for providing dimensional or positional data "on the fly" in a
highly accurate manner and at rates which are extremely fast, particularly
as compared with contact-type robot sensors of the prior art. Generally
speaking, the invention involves the use, in robotic apparatus employing
optical signals of light pulsing for "freezing" robot position
determination data in time. For example, the invention is applicable to
systems which use X-, Y-, and Z- axis optical encoders and like optical
encoders, optical sensors and/or optical axis compensators and in
operation, the various devices are pulsed or strobed simultaneously so as
to "freeze" the readings of the electro-optical detector arrays or optical
readers. All of these readings are then fed to a computer which accurately
computes the part location based on these input data.
In one embodiment, at least one encoder is used in determining the position
of the robotic apparatus and the optical reader is triggered by a pulse
from the encoder. Typically, a plurality of encoders are used and the
reader is triggered from one or more selected encoders, e.g., the encoder
experiencing the greatest rate of change, i.e., that undergoing the
greatest relative movement or that detecting the greatest dimensional
change in a part under inspection. The light source of the encoder can
also be pulsed to determine the absolute position of the encoder at the
time of the pulse. Further, internal and external axis compensation
devices can be pulsed simultaneously.
Other aspects of the invention concern the provision of positional data for
robotic apparatus through the use of several different techniques. In one
embodiment, a scanning detector array is utilized in inspection of a
continuous surface and is scanned at a very high rate so as to "blur" the
data output therefrom, curve fitting techniques being used to reconstruct
the surface contour.
According to a further aspect of the invention, optical linear or rotary
encoder apparatus is provided which, in one embodiment, uses a scale
employing widely spaced lines together with a combination continuous wave
and pulsed light readout, or with a pulsed diode array readout only. In a
related embodiment, the teeth on the rack of a rack and pinion drive
mechanism, the threads of a drive screw, and the like can be read rather
than an encoder scale. In another embodiment, a conventional scale with
relatively coarse spacing is used in combination with a pulsed light
source diode array readout.
Other features and advantages of the invention will be set forth in, or
apparent from, the detailed description of the preferred embodiments which
follows.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is schematic, perspective view of robotic apparatus of one type to
which the present invention is applicable;
FIG. 2 is a side elevational view of a detail of the sensor of FIG. 1;
FIG. 3 is a flow chart outlining the basic operation of the control system
of the invention;
FIG. 4 is a schematic diagram of an encoder arrangement constructed in
accordance with a further aspect of the invention;
FIG. 5 is a schematic diagram, in perspective, of a linear encoder
constructed in accordance with a further embodiment of the invention;
FIG. 6 is schematic diagram, in perspective, of a further encoder
embodiment of the invention; and
FIG. 7 is a schematic diagram, in perspective, of a rotary encoder
embodiment of the invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring to FIG. 1, a system is illustrated which is used to provide
position data or to correct position data on a coordinate measuring
machine (CMM) or cartesian axis robot. The system is of the general type
disclosed in U.S. patent application Ser. No. 262,492 referred to above
and reference is made to that application for a description of the details
thereof. It would be understood that the illustrated system is merely
exemplary of the type of system to which the invention is applicable and
that the invention is not limited to application to such a system. The
system illustrated in FIG. 1 includes a base 10 along which a motorized
vertical (Y-axis) arm 12 travels in the direction of the Z axis. A
horizontal arm 14 mounted on vertical arm 12 moves in and out along the
X-axis. Further rotational axes, e.g., rotational (.theta.) and yaw
(.phi.), are provided at the end (wrist) of horizontal arm 14 which enable
a sensor package or unit 16 located at the end of arm to rotate about the
X-axis and about the yaw axis under the control of a suitable drive unit
(not shown). A sixth wrist roll axis, and/or part rotation axis (not
shown) can also be provided.
An X-axis, horizontal arm encoder 18, a Y-axis vertical arm encoder 20 and
a Z-axis, base encoder 22 are also provided for indicating the X, Y and Z
coordinates of sensor 16, while a pair of rotational encoders 24 and 26
indicate the angular portion. The output signals from all of the encoders
are fed to computer (not shown).
In addition, as disclosed in Ser. No. 262,492, a plurality of axis
compensation (AC) sensors are provided. Considering the sensor arrangement
which provides X-axis compensation as exemplary, a laser 28 produces a
beam 30 which is received by a position sensitive detector 32 comprising a
driftless matrix array or analog sensor, e.g. an UDT-SC-10. (It is noted
that the beam can be located higher up on arm 12 as indicated by beam 34
and detector 36.) A similar arrangement of a laser 38 and detector 40 is
provided for correction of "droop" or downward Y-axis deviation of
horizontal arm 14. It is noted that an additional arrangement (not shown)
can be mounted on vertical arm 12 and that other sensors can be employed,
such as the external axis compensation sensors disclosed in Ser. No.
262,492. Further, optical axis compensation sensors modeled after the
displacement sensing devices disclosed in U.S. Pat. No. 4,276,698 (Dore'
et al) can also be used.
In the illustrated embodiment, the optical sensor unit 16 is used to sense
the contour of a part 42 (e.g. an automobile hood) mounted on a turntable
44.
As shown in FIG. 2, optical sensor unit 16 comprise an optical
triangulation system comprising a light source 45 such as a pulsed diode
laser and a lens 47 which focusses an image spot onto a detector array or
optical reader 49, such as a Reticon 256C.
Turning now to the present invention, as discussed above, measurement speed
is critical in inspection systems of the type being considered and it is
the aim of the invention to provide accurate inspection of parts
"on-the-fly", in contrast to the contact feeler inspection systems of the
prior art. In general, the present invention involves reading,
electronically, the output data from all dimension sensors at the same
time, i.e., considering the system of FIG. 1, simulataneous reading of the
outputs of the X-, Y- and Z-axis encoders, diode array sensors, and axis
compensators, etc., and in this way, to compensate , inter alia, for droop
of the horizontal arm 12 as well as for inertial forces causing unwanted
deflections of the sensor unit 16. This is accomplished in accordance with
the invention by "flashing" the light source of the diode arrays, i.e.,
strobing the array with a short light pulse to "freeze" the triangulation
spot image (or other image such as a profile or reflected image) when
triggered by an encoder pulse or to trigger the "read" cycle of the
optical sensor (detector) array from the axis encoder which is moving the
fastest or in the direction of greatest sensor signal (the increase in
sensor signal being due for example, to a rapid change in the shape of the
part owing to a steep slope in the surface contour). Advantageously, for
true three (or more) axes measurement all three (or more) encoder readings
are "freezed" simultaneously with each other and with the outputs of the
sensors. As explained below in connection with FIG. 4, a pulsed laser
system can be used for this purpose. Further, if axis compensation sensors
are used, the readings thereof should also be "freezed" in response to the
read command signal.
Referring to FIG. 3, a flow chart for pulsed readout system is shown. As
illustrated, a "read data" signal is commanded which results in the
issuance of a pulse command to the various sensors illustrated in FIG. 1,
including encoders and optical axis compensation (AC) arrangements. The
corresponding detector arrays are read simultaneously and the resultant
outputs are fed to a computer which computes the part location based on
the data received thereby.
Referring to FIG. 4, a pulsed light source encoder arrangement is shown.
One input to a computer 50 is provided by a light source 52 which focuses
a light beam through a scale 54 onto a conventional readout device 56. In
addition, a diode laser and associated beam expander 58 is provided which
produces a laser pulse in the form of, e.g., a 100 n sec burst which is
transmitted through scale 54 and a lens 60 to a detector array 62. The
diode array axis and the conventional sensor encoder axis (e.g., phase
quadrature) are geometrically related, and the techniques disclosed apply
to rotary as well as linear encoders. Through the provision of a fast
pulse light source the encoder arrangement of FIG. 4 is capable of
"on-the-fly", rapid, submicron measurements. A somewhat similar
arrangement is disclosed in U.S. Ser. No. 134,465 (see pages 14 to 18 and
FIG. 12A), now U.S. Pat. No. 4,403,860, issued on Sept. 13, 1983. Further,
one type of scale that can be used as scale 54 in FIG. 4 is that disclosed
in U.S. Pat. No. 4,074,258.
However, in accordance with a preferred embodiment, scale 54 comprises a
conventional scale made up of a plurality of equispaced lines. The
photodiode detector array 62 is able to determine, in an absolute manner,
the shift of lines at the strobed time of the array, in manner similar to
that disclosed in Ser. No. 134,465, referred to above. This approach
permits a very coarse scale to be used, i.e., one in which the spacing
between lines is relatively wide, (e.g., 0.002 inch) while still achieving
very high resolution. This high resolution can be provided using the
techniques disclosed in U.S. patent application Ser. No. 163,290, now U.S.
Pat. No. 4,394,683, issued on July 19, 1983, whereby the scale divisions,
i.e., the spacings between lines, can, in effect, be broken down into
smaller divisions using the digital interpolation techniques described in
the application in question. In fact, because of this high resolution of
the sensor array output, the scale used can be, relatively speaking, very
coarse, i.e., lines every 0.05 inch, to provide readings to 10
microinches. In such an embodiment, the conventional encoder sensor
arrangement can be eliminated because data can be taken at a rate fast
enough to eliminate missed data due to movement of scale 54 more than 0.05
inch between scans.
To explain, reference is made to FIG. 5, which discloses an encoder system
including a pulsed light source 70 which directs light pulses past a scale
72 having scale lines spaced, in this example, 0.05 inches apart and a
lens 74, providing 4:1 magnification, to a detector array 76 typica11y
consisting of 1000 detectors spaced 0.001 inches apart. A signal
processing circuit 78 is connected to the output of array 76 and the
arrangement provides at least 35.times. magnification. If the scanning
rate is 500 scans per second so that a scan is provided every 0.002
second, then for a line to move on and off in one scan (thereby losing
track), the scale 72 would have to move 500.times.0.05 or 25 inches per
second which is considerably faster than is required for the system. A
computer 80 connected to the array enables the system to start at the
"home" position line and to provide appropriate scanning.
A further, low cost, coarse reading embodiment of this aspect of the
invention is that illustrated in FIG. 6. In this embodiment, the light
output beam from a light source 90 impinges on the teeth 92 of the rack 94
of a rack and pinion drive mechanism 96. A pinion 98 is connected to a
drive motor 100 whose rotation is monitored by a rotational encoder 102.
The beam from source 90 is focused by a lens 104 on a linear array 106
which determines the relative position of the image of rack teeth 92. This
arrangement eliminates an encoder as such, and reads the position of the
rack teeth 92 to upgrade the data from the rotational encoder 100 of drive
motor 98. This approach eliminates errors caused by backlash and the like,
and by averaging the reading over a large number of teeth 92, the effect
of local gear errors can be obviated. It will be understood that this same
approach can be used in a ball screw drive, and like mechanisms.
It is noted that the systems of FIGS. 4 to 6 can be factory calibrated
using a laser interferometer or the like, a process which is especially
easy with systems using coarse spacing or pitch. The use of high
resolution linear array sensor, whether pulsed or not, permits a degree of
interpolation between the scale lines such that coarse spacing can be used
so that it is necessary to store only a few, relatively speaking,
calibration points. For example, 1000 calibration points can be stored on
0.050 inch centers for a 50 inch scale.
It is also noted that while the encoders discussed above use linear scales,
a rotary scale can also be used as is illustrated in FIG. 7. In FIG. 7, a
light source 110 provides imaging of the peripheral scale lines 112 of a
rotary disc member 114 onto a detector array 116 via a lens 118. Optical
fibers (not shown) can be employed to enable detector 116 to be remotely
located.
To recapitulate, and expand the possibilities with respect to providing
on-the-fly readings, techniques that can be used include strobing the scan
of the diode array or pulsing the light source used for illumination with
a very short burst as the encoder moves from one count to the next,
because this point on the encoder can be well defined as to the position
thereof. Another approach involves operating the diode array at an
extremely high scan rate or short integration time relative to the robot
part movement rates to avoid blur (an approach which requires a large
amount of light power which may or may not be available). A further
technique that can be used with continuous surfaces is to employ a curve
fitting operation to reconstruct the original curve in an acceptable
manner even in the presence of "blurred" data. An additional technique
which is usable even with single encoder systems but is more necessary
with multiple encoder systems is to utilize pulsed light sources to freeze
the interpolated encoder portions at the same time a "freeze motion" light
source is used in conjunction with optical sensor so that both the encoder
and sensor data are frozen in synchronism.
It should be noted that the sensors used can be of the type disclosed in my
copending U.S. patent application Ser. No. 461,685, entitled "Sensing
Location of an Object", filed on Jan. 27, 1983, which application is
hereby incorporated by reference.
It is also envisioned that high speed light sensing will be coupled with
the data acquisition from the various encoders and axis compensation units
described above. Thus, it is within the purview of the invention to
provide simultaneous reading of all the encoders, the axis compensation
units and the sensor unit, whether or not any, or all, of the light
sources used therewith are actually pulsed.
It is noted that the axis compensation provided, which typically can be
carried out as disclosed herein with a matrix photodetector array,
benefits from the use of a pulsed light source since the array scan rate
is typically not fast enough to measure rapidly the position of the
mechanical structure in two axes. If one axis (linear) detector arrays are
used, that is, those providing compensation in only one axis, these arrays
can be much faster in operation.
Accordingly, it is also within the purview of this invention to transmit a
trigger pulse to the light source of the sensor unit, for example, and
thereby cause that unit to produce a light pulse to freeze sensor data. In
turn, a strobe pulse is transmitted to lock up the readings of all the
axes encoders as rapidly as possible. Obviously, if the axes compensation
units are to be read out with matrix arrays, it would also be highly
advantageous to provide for "firing" a light pulse in the axes
compensation sensors at the same time that "firing" the light pulse in the
sensor unit takes place, thereby freezing their readings as well. If
conventional axes encoders (i.e., optical encoders, inductosyn encoders
and the like) are used, one would read the last encoder count before the
strobe pulse to the encoder is received. Naturally, if pulsed light source
encoders such as disclosed above are used, these encoders could also be
pulsed.
In many cases, however, it is not possible to effectively pulse the light
source of the sensor unit with a short burst sufficient to "freeze" part
data. This has to do with certain power level requirements that must be
met to operate at least with pulse light sources presently available as
well as with array sensitivities, and considerations involving the form,
reflectivity, color and condition of the part surfaces, and the like, from
which light must be returned to the sensor.
Our solution to such problems is to use an image intensifier to amplify the
faint light reflected from the surface within a short weak pulse. However,
this approach also amplifies other ambient light as well, which may be a
problem in factories. An image intensifier, such as a microchannel
intensifier, can itself be time rated, so that rather than having to pulse
the light source, one can "pulse" the sensor.
Another approach in dealing with the problem of long integration times is
to provide another mode of machine operation, using the system illustrated
in FIG. 8, in which a start read pulse is sent to the sensor unit from a
central controller or host computer. This pulse can further be produced,
but is not necessarily produced, in response to a start signal from one of
the moving coordinate axes, typically the one in which the greatest
changes are noted.
On receipt of the start signal, the sensor unit detector array (or other
sensor unit) commences its integration of the light received from the part
and, in accordance with an important difference from the previously
described embodiments, the machine encoders continue to read during this
integration period until such time as the sensor signals back that it has
received enough light to get a reading. At this time a "finish read"
signal is produced. In addition, the positions of the encoders at the end
of the reading are noted and the mean (or other) position over which the
data is taken is computed from the start and stop positions of the
encoders. This mean position is assumed to correspond to the mean location
of a spot, line or edge on the sensor array during the integration time
(during which a slight blur, a spot, or the like occurs).
As has been discussed above, matrix arrays of the type being considered
typically require relatively long integration times in order to receive
sufficient light back from the test surface, from either the white light
surface illumination or the diode laser triangulation projection, in order
to obtain enough data to produce a meaningful reading. Typically, such
integration times can be on the order of 1/10 to 1/200 of a second. While
these times may appear fast, such times are actually relatively slow if
the measuring machine arm is moving relatively rapidly over a part,
particularly one whose contour is also changing rapidly. Accordingly, to
produce a measurement in this case, as discussed above, an external data
command from a host computer is given to the sensor controller to start
scanning. At the same time, a data request is made to the axes encoders
and axes compensation units (if used) for their positional data at the
start of the detector array integration.
When the sensor has returned a "finish read" command to its controller, the
final positional data from the axes encoders and axes compensation units
(if used) are requested and this data transferred to the computer. By
knowing the starting and end points of the axes encoders and axes
compensation (if used) during the measurement, the mean location in space
of the sensor during the measurement can be obtained. This location
information, when combined with part location data produced by the sensor
unit, provides the data as to where the part is in space.
In a case where both integration and scan time is appreciable in the
operating mode discussed above, the "finish read" signal should be
produced after the scan is complete. If the integration time is short
relative to scan time, the "finish read" signal should be produced after
the integration time is complete and the light source turned down (to
thereby stop integrating the light charge on the sensor array).
Another aspect of this process should also be considered. Specifically,
while the integration referred to above is taking place, the spot width or
image is moving on the detector array, due to the motion of the machine.
In the method described, the mean location is the same regardless of the
direction of the scan, so long as the array can be read out at high speed
(e.g. 1/2000 sec.) even with a long integration time (e.g. 1/30 sec.).
This is the case with many linear arrays. However, if the scan time is
also long, the direction of scan relative to the array will make a
difference with respect to the spot or edge position. Compression results,
for example, when the scanning direction of the array is opposite to that
of the movement of the "spot" (or line or the like) on the array surface.
Conversely, an elongation of the spot, and a shift in the apparent
position thereof, will occur if the scan is long, and in the same
direction, as the spot movement.
When scan rates are relatively long, blur errors can advantageously be
minimized by scanning the array (or like devices, e.g., TV camera, etc.)
in the direction opposite to the direction of movement.
Despite the emphasis on pulsed light sources in the discussion above, the
invention is also applicable for use with CW sources and high speed array
scans and/or data strobes to "lock up" co-ordinate axes and/or
compensation data at high speeds.
The use of curve fitting has been described above as an approach in
reducing blur error at slow scan rates for relatively continuous contour
surfaces where array data points can be taken. Another axes interpretation
technique, usable even with one scan only, will now be described.
Suppose the photodetector array sensor, such as that illustrated in FIG. 2,
requires 1/30 of a second to integrate enough light on the detector array
to provide a good reading from a surface with a relatively low
reflectivity, even with the maximum light energy that can be produced by
light source 45. To substantially reduce or eliminate blur error it is
necessary to: (1) slow the axes rates down to where, in 1/30 of a second,
negligible movement occurs, or compensation is required, or (2) determine
the mean position of the axes (and axes compensation units if used) during
the sensor integration time, and assume that the mean position sensed
(which, in the embodiment of FIG. 2, would be a triangular spot on the
array), during the integration time, occurs at the mean axes position
during the 1/30 second period.
Clearly the second option allows faster reading rates, although both
options can be combined.
We will now consider another approach in utilizing relatively slow scan
matrix array devices with long integration times compatible with
conventional light sources. A serious problem with matrix arrays is the
long time period necessary to read all the elements of the array. For
example, even at a fast 1 MHZ clock rate a 256.times.256 element matrix
array having approximately 64,000 elements takes 1/15 second to fully read
out. This is a long time for a sensor mounted on a moving machine (or
sensing a moving part) if high accuracy is desired.
For triangulation use however, as in the embodiment of FIG. 2, one can
actually use the sensing of the spot to control the axes encoder data
acquisition and possibly the sensor array clock rate. For example, if the
array 49 in FIG. 2 is a matrix array, the charge is first dumped from the
array before measurement begins. At this point, the approximate location
of the projected spot (line, edge etc.) on the array is noted. Then, on
the next scan, it is only just when the approximate zone of the spot is
detected that the coordinate axes encoder data starting point are enabled
and only just after the spot is scanned by the array, when the encoder
data end points are locked up, in order to obtain the mean. Thus, for
spots which occupy only 20 lines of the array, the reduction in reading
time is 25:1 on a 256 element array.
In the same vein, one can, in addition, control the clock rate of the
array, by, for example, slowing it down (and thus allowing more
integration time) only in the zone of the array where the spot appears.
While usable primarily in connection with triangulation spots, this same
technique can also be used on reflected or transmitted waves to decrease
blur where a zone of reading can be determined.
The technique just disclosed for locking up the axes encoders at the start
and stop points of the zone can also be applied to the start and stop
points of an axes compensation sensor.
In a more sophisticated version of the FIG. 8 control system, the total
"frame" of the array, i.e., substantially all elements, can be read into a
memory at a given clock rate, while reading all encoder data from axes
encoders (and axes compensation units, if used) into a second memory in
timed relationship with respect to the array clock rate. Then, after the
fact, the triangulation spot, edge image, or any other feature, can be
found by scanning the image stored in the first memory, and when found,
its position can be noted relative to the encoder positions stored in
memory to determine the encoder values at the position in time that the
image point or feature of interest was determined.
Although the invention has been described in detail with respect to
exemplary embodiments thereof, it will be understood by those of ordinary
skill in the art that variations and modifications may be effected within
the scope and spirit of the invention.
* * * * *
|
|
 |
|
 |
Description |
|
